[Feature]: Stop the job gracefully in case of a spot instance interruption

[peterschmidt85]

Problem

Currently, if a spot instance is interrupted, the job is killed abruptly.

Solution

AFAIK, AWS/GCP/Azure provides mechanisms to recieved events about the interruption. If dstack-shim is able to get such an event, it can initiate a graceful stop of the job (e.g. by sending SIGINT to the container).

Workaround

No response

Would you like to help us implement this feature by sending a PR?

Yes